5. During the evaluation of two speech recognition algorithms, A1 and A2, each is presented with the same sequence, u1, . . . , un, of labeled utterances for recognition.

The {ui} are assumed to be a random sample from some population of utterances. Each algorithm makes a decision about the label of each ui which is either correct or incorrect. The data are shown below:

Is there evidence that the two algorithms have different error rates? Test at α = 0.05 using both the paired T test procedure and McNemar’s test procedure.

Correct A2 Incorrect Correct 1325 3 A1 Incorrect 13 59